2D CONTOUR: FREE ENERGY SURFACES (xy plane)

• Checking convergence of fes over the last ~20ns of simulations

FES:most recent

with walker2@90 (running) , walker3 failed@60, walker4@30 (running w/ sander, walker1@60 (walkers0,5,6,7,8 complete@100ns)

• pore minima has gotten significantly deeper (126 vs 116 previously)

(<Figure size 720x720 with 2 Axes>,
 <PolarAxesSubplot:>,
 <matplotlib.contour.QuadContourSet at 0x7ff7a8c0a370>)

FES100

• after walkers5,6,7,8 completed 100ns
• appearance of new minimum protruding into DIII extending from the middle of DIII-IV fenestration

(<Figure size 720x720 with 2 Axes>,
 <PolarAxesSubplot:>,
 <matplotlib.contour.QuadContourSet at 0x7ff7d8107910>)

FES90

DIII-IV well sampled, up to the wall restraint boundary

(<Figure size 720x720 with 2 Axes>,
 <PolarAxesSubplot:>,
 <matplotlib.contour.QuadContourSet at 0x7ff7a24d0d90>)

FES80

(<Figure size 720x720 with 2 Axes>,
 <PolarAxesSubplot:>,
 <matplotlib.contour.QuadContourSet at 0x7ff7d0c78310>)

PLOTTING HILL HEIGHTS

Timeline of Hills Deposited by All Walkers

• Plot of Hill heights vs Wallclock time (of all walkers combined)

Text(1616804314, 1.2, 'FES100 Written')

Checking hill heights of walkers that completed 100ns: #7,8,6,5,0 (in order of which finished first)

• Walker7 is staying exclusively within the pore cavity for the last 20ns, which appears to be in a well-sampled area, whereas Walker8 is exploring DIII-IV and is still depositing fairly large hills. Therefore, could keep running past 100ns until walker8 hill heights look like walker7...

Checking hill heights of incomplete walkers: #1,2,3,4 that failed at various timepoints

• Given the reduced heights of these problematic walkers, perhaps not worth trying to restart them.

<AxesSubplot:xlabel='time', ylabel='heights'>

SAMPLING OF PORE SPACE: X Y Z

z and theta variables transformed to xy coordinates using:

• x = z * sin(theta) + pi/4
• y = z * cos(theta) + pi/4

in order to check sampling into each fenestration.

Harmonic restraints placed on:

• s @ 12 and -12 A (upper and lower boundary along pore axis)
• z @ radius of 18 A (circular boundary along xy plane)

Complete (100ns) walkers: 0,5,6,7,8

• extensive sampling into fenestration DI-II by walker6
• A single walker tends to explore into each fenestration, hit the wall restraints, and then return to the pore ( Walker0->DII-III; Walker6->DI-II; Walker8->DIII-IV ). Subsequent walkers to enter each fenestration tend to not traverse as far.
• Intracellular pore opening well-sampled by Walkers0&5

Text(50000, -17, 'Pore Gate')

Incomplete walkers: 1,2,3,4

FAILED WALKERS:

Walker	Inital Fail	Restarts	Notes
0	72ns	Restarted from 70	COMPLETED to 100ns
1	65ns	Restarted from 60	Multiple restarts (tried CPU instead of GPU) - did not get past 1ns
2	43ns	Restarted from 40; 60; 70	..still running
3	27ns	Restarted from 20; 60; 60	last fail at 65
4	13ns	Restarted from 10; 20; 30	still running v slowly (sander)

• Lagging walkers have not sampled as far into each fenestration
• Walkers1-4 seem highly mobile and combined with the fact that heights of hill depositions is significantly reduced, suggests that the pore and fenestration space has been adequately sampled and that the energy surface has been sufficiently flattened.
• Movement of Walker1 is confined within the pore cavity after 60ns as it repeatedly failed without writing a new restart file and was restarted using the old rst7 (hills were kept as they contributed to the metad potential of other walkers)

Text(55000, -9.5, 'Pore Gate')